Anthropic rejects Department of Defense request to lift security controls on AI
Anthropic has rejected the Pentagon’s request to lift its AI safeguards after the Pentagon warned that the company’s contract could be at risk.
NEW YORK – Anthropic filed a lawsuit on Monday, March 9, to block the Pentagon’s national security blacklisting, escalating a high-stakes battle with the U.S. military over limits on the use of the artificial intelligence lab and its technology.
Anthropic said in its lawsuit that the designation is illegal and violates its free speech and due process rights. The filing in federal court in California asked a judge to revoke the designation and block federal agencies from enforcing it.
“These actions are unprecedented and illegal. The Constitution does not allow the government to use vast powers to punish companies for protected speech,” Antropik said.
The Department of Defense granted Anthropic a formal supply chain risk designation on Thursday, March 5, restricting its use of technology that officials say is used in military operations in Iran. Pentagon Secretary Pete Hegseth named Anthropic after the startup refused to remove guardrails on the use of AI for autonomous weapons and domestic surveillance. The two countries have been in increasingly heated debates over these restrictions for months, Reuters first reported.
Antropic officials said the lawsuit does not prevent them from resuming negotiations and reaching a settlement with the U.S. government. The company says it has no intention of fighting the government. The Pentagon said it does not comment on litigation. Last week, Pentagon officials said the two countries were no longer in active talks.
The designation poses a major threat to Anthropic’s business with governments, and the outcome could affect how other AI companies negotiate limits on military uses of their technology. The company’s CEO, Dario Amodei, clarified on Thursday, March 5, that the designation is “narrow scoped” and that companies will still be able to use the tool on projects unrelated to the Department of Defense.
President Donald Trump also told the government to stop working with Anthropic, which financially backs companies such as Alphabet Inc.’s Google and Amazon.com Inc. Mr. Trump and Mr. Hegseth said they planned for a six-month phase-out.
Reuters reported that Anthropic investors are racing to contain the damage from the conflict with the Pentagon.
Mr. Trump and Mr. Hegseth’s Feb. 27 actions came shortly after Mr. Amodei met with Mr. Hegseth in hopes of reaching an agreement after months of discussions with Anthropic over whether the company’s policies could constrain military action.
The Pentagon said U.S. law, not private companies, determines how to protect the country, insisted it would have complete flexibility in using AI for “all lawful uses,” and argued that Anthropic’s restrictions could put American lives at risk.
Anthropic said that even the best AI models are not reliable enough to be fully autonomous weapons and are dangerous to use for that purpose. The company also drew a line under the surveillance of Americans in the country, saying it was a violation of fundamental rights.
After Hegseth’s announcement, Anthropic said in a statement that the designation is legally unsound and sets a dangerous precedent for companies negotiating with the government. The company said it would not waver from “threats and penalties,” and on Thursday, Amodei reiterated that Anthropic will challenge the designation in court.
He also apologized for an internal memo published by tech news site The Information on Wednesday. In a memo written last Friday, Amodei said one reason Pentagon officials don’t like the company is that “we don’t admire Trump like a dictator.”
Last year, the Department of Defense signed contracts worth up to $200 million each with major AI labs, including Anthropic, OpenAI, and Google.
Microsoft-backed OpenAI announced a deal to use its technology on Department of Defense networks shortly after Hegseth was added to Anthropic’s blacklist. CEO Sam Altman said the Pentagon shares OpenAI’s principles of ensuring human oversight of weapons systems and opposing mass surveillance by the United States.
